Proc. Nadl. Acad. Sci. USA Vol. 86, pp. 7128-7132, September 1989 Two targeted to the same lytic granule compartment undergo very different posttranslational processing (cytolysin/Na-benzyloxycarbonyl-L-lysine thiobenzyl estrase/immunoelectron microscopy/N-glycosylation/mannose 6-phosphate) JANIS K. BURKHARDT, SUSAN HESTER, AND YAIR ARGON* Department of Microbiology and Immunology, Duke University Medical Center, Box 3010, Durham NC 27710 Communicated by D. Bernard Amos, June 19, 1989

ABSTRACT The granules of natural killer (NK) cells membranes of target cells after exposure to cytotoxic lym- contain cytolysin and serine proteases, proteins that are ex- phocytes (5). Also packaged in thegranules is afamily ofserine pressed specifically in cytolytic cells and are released in re- proteases (6), often called granzymes (7), whose expression is sponse to target binding. We have used immunofluorescence restricted to CTLs and NK cells (6-8). In rat NK cells, two of and immunoelectron microscopy to localize these proteins with these proteases have been classified as Na-benzyloxycar- respect to the various morphological compartments ofgranules bonyl-L-lysine thiobenzyl esterases based on their substrate in a rat NK line. Both cytolysin and the proteases are specificity (9). At least one ofthe two, granzyme A, is secreted limited to the core regions of the dense core granules. While during cytolysis (10), but its exact function is not yet known. these proteins are targeted to the same compartment, they A third component of NK granules is a chondroitin sulfate differ markedly in their posttranslational processing. Cytolysin proteoglycan, which is released during the killing process (11). bears N-linked oligosaccharides that are converted to the The presence ofproteoglycans is typical of secretory granules complex type, while the major trypsin-like protease, granzyme in various cell types (3). A, bears only high-mannose-type oligosaccharides. The glycans Although the lytic granules function as secretory granules, of granzyme A, but not those of cytolysin, are modified with they are also similar in many ways to lysosomes. Like phosphomannose moieties. These results suggest that one pos- lysosomes, the granules are acidic (9), although this is also a sible mechanism for packaging proteins into NK granules is the property of secretory granules (3). In addition, NK granules mannose 6-phosphate-dependent lysosomal targeting system. contain many "classical" lysosomal hydrolases, including However, the absence ofthe mannose 6-phosphate modification acid phosphatase, f3-glucuronidase, and aryl sulfatase (12- from cytolysin suggests the existence of yet another targeting 14). Indeed, histochemical assays ofthese enzymes are often system. used to identify large granular . Interestingly, at least some of the lysosomal enzymes are secreted during cytolysis (14, 15), together with the granule-specific proteins. Lytic granules are specialized organelles found in cytolytic This observation implies that the usual sorting of secretory lymphocytes such as cytotoxic T lymphocytes (CTLs), nat- and lysosomal proteins into separate organelles is incomplete ural killer (NK) cells, and lymphokine-activated killer cells. in cytolytic cells, either because the same organelle mediates The binding ofan appropriate target cell, such as a tumor cell both functions or because the population of lytic granules is or a virally infected cell, triggers the directed ofthe actually composed of two distinct organelles. granule contents (1). There is now considerable evidence that Electron microscopic (EM) studies of NK granules show these granule contents can effect the subsequent of the that they are, in fact, morphologically heterogeneous (16). target cell. Although the granules clearly have a specialized Like the condensing vacuoles seen in other secretory cells, secretory role, they also have properties in common with many NK granules contain a homogeneous electron-dense lysosomes. This dual nature raises interesting questions core. Granules having a dense core surrounded by a thin about the biogenesis of lytic granules and the routing of cortex containing membranous material were classified by proteins to them. Neighbour et al. (16) as type I. A second morphological class Granule exocytosis in NK cells is a process of regulated of granules [type II (16)] is larger, more irregularly shaped, and polarized secretion. The Mg2"-dependent binding of NK and resembles multivesicular bodies. Often, a mixed granule cells to targets establishes an axis of polarity through the morphology is observed, where a type I dense core is killer cell and induces a substantial remodeling of its archi- included within a large type II region. tecture (2). This remodeling involves the reorientation of the The relationship among the three types of granules is not Golgi apparatus, cytoskeletal elements, and lytic granules understood. It is not known if the morphological types repre- toward the bound target. The granules then fuse with the sent a maturation sequence of lytic granules. Moreover, it is plasma membrane domain that contacts the target cell and not clear which granules are actually used in cytolysis. EM release their contents. Both the reorientation and the subse- images have been interpreted (16) to show that both type I and quent exocytosis require Ca2", a typical requirement for type II granules can secrete their contents, but this has not regulated secretion (3). been directly demonstrated. Since no nongranule organelle Packaged in the lytic granules are proteins whose expression has been identified as a lysosome in NK cells, it is quite is specific to cytolytic lymphocytes and which are secreted in possible that one type ofgranule corresponds to the secretory response to target binding. Foremost among them is cytolysin lytic granule, while the other type fulfils lysosomal functions. (4), which is analogous to perforin (5). It is thought to effect To address the relationship between lytic granules and target lysis by a Ca2+-dependent oligomerization into cylin- lysosomes, several granule proteins have been localized at drical amphipathic pore structures (1, 5). These pores are the EM level. Perforin was found in the dense core of type I capable of lysing cells and resemble pores observed in the Abbreviations: CTL, cytotoxic T ; EM, electron micros- The publication costs of this article were defrayed in part by page charge copy; endo H, endoglycosidase H; NK, natural killer; SBTI, soybean payment. This article must therefore be hereby marked "advertisement" trypsin inhibitor. in accordance with 18 U.S.C. §1734 solely to indicate this fact. *To whom reprint requests should be addressed.

7128 Downloaded by guest on September 25, 2021 Immunology: Burkhardt et al. Proc. Natl. Acad. Sci. USA 86 (1989) 7129 granules of human CTLs (17). Recently, serine proteases Analysis ofCarboydrates. Serine proteases labeled with 32Pi were also shown to reside in similar granules ofmurine CTLs were immunoisolated from RNK-16 cells by affinity chro- (18). Sulfated proteoglycans have also been localized to type matography on SBTI agarose (Sigma) essentially as de- I granules, most likely to the dense core (11). In contrast, scribed above. The isolated material yielded a single labeled several hydrolytic enzymes are excluded from the dense core band on NaDodSO4/PAGE, which comigrated with the ma- and are localized in the cortical region of type I as well as in jor of the anti-granzyme A (data not shown). After type II granules (12-14). Taken together, these observations digestion to completion with endo H, the samples were suggest biochemical as well as morphological compartmen- chromatographed on a Bio-Gel Pa (200-400 mesh) gel fil- talization of lytic granules. tration column (120 x 1 cm) in 50 mM ammonium acetate (pH To understand this compartmentalization better, we have 7.0). The column was calibrated with 14C-labeled purified examined the biosynthesis of two cytolytic lymphocyte- oligosaccharides (gift of M. Snider, Case Western Reserve specific proteins, cytolysin and granzyme A. In particular, University) and with nitroiodophenylacetyl. The excluded we asked if they behave like lysosomal proteins or like and included volumes were determined by using 1251-labeled secretory granule proteins. Moreover, we asked if the two bovine serum albumin and [3H]mannose, respectively. The proteins are targeted to the same or different morphological peak fractions were pooled, lyophilized, resuspended in regions of granules in NK cells. water, and subjected to mild acid hydrolysis, catalyzed by Dowex 50 H' resin (24), at 970C for 16 hr. The sample was MATERIALS AND METHODS dried quickly with an Evapomix apparatus (Buchler Instru- ments) and resuspended in water. The resulting monosac- Cell Culture and Labeling. RNK-16 clone CRC- is an in charides were analyzed by thin-layerchromatography. What- vitro-adapted clone derived from a large granular lymphocytic man K6 silicagel plates were loaded with radioactive samples leukemia that arose spontaneously in aging Fisher rats (19) mixed with individual cold monosaccharides and developed (gift of C. Reynolds, National Cancer Institute, Frederick, with one of two solvents. Solvent A (2-propanol/water, 4:1) MD). This clone retains NK characteristics, including mor- enabled separation of mannose 6-phosphate from glucose phology, surface markers, and killing specificity, and does not 6-phosphate and N-acetylglucosamine 1-phosphate. Solvent require or respond to interleukin-2 (19). The cells were grown B (ethylacetate/pyridine/acetic acid/water, 5:5:1:3) re- in RPMI 1640 medium supplemented with nonessential amino solved mannose 6-phosphate and glucose 6-phosphate from acids, pyruvate, glutamine, and penicillin/streptomycin, and N-acetylglucosamine, mannose, glucose, and amino sugars. containing 5% fetal calf serum. Cells were labeled metaboli- After chromatography, the plates were dried, exposed to cally at a density of 5 x 106 cells per ml, as described (20). film, and the positions of the cold monosaccharides were [35S]Methionine labeling (15 mCi/ml; 1444 Ci/mmol; 1 Ci = 37 detected by charring with H2SO4. GBq; Amersham) was at 100-250 ,uCi per ml for the indicated times. For labeling with [9,10-3H]palmitate (1 mCi/mi; 54 RESULTS Ci/mmol; New England Nuclear), cells were incubated at 150 .uCi per ml for 4-6 hr. For labeling with 32p; (294 mCi/ml; 8835 Cytolysin and Serine Proteases Reside in Type I Granules. Ci/mmol; ICN), cells were labeled with 1 mCi/ml for 3-4 hr Although cytolysin and the serine proteases have been pu- in phosphate-free medium. rified from NK granules, direct microscopic demonstration Immunofluorescence and Immunoelectron Microscopy. that they in fact reside in the same organelle has not been Monospecific rabbit anti-cytolysin antiserum was prepared provided. This becomes even more important when one by P. Henkart (National Cancer Institute, Bethesda, MD). considers the morphological heterogeneity of lytic granules. This detects a single band in immunoblots and can The ultrastructure of RNK-16 cells is illustrated in Fig. 1. inhibit lysis of erythrocytes by purified cytolysin (21). For These cells, like other large granular lymphocytes, contain microscopic detection of serine proteases, we used soybean numerous type I granules, with a uniformly electron-dense trypsin inhibitor (SBTI; Sigma), which was biotinylated core, surrounded by a thin cortex ofmembrane lamellae. The according to ref. 22 at a biotin/ molar ratio of20. This type II granules of RNK-16 are much larger and more reagent retained its full activity when tested against trypsin, and its binding was completely inhibited by excess unconju- gated SBTI. Biotin-SBTI was detected with streptavidin- 4,. Texas red (Vector) or streptavidin-colloidal gold (Janssen). Immunofluorescence microscopy was performed with a Zeiss Axiophot microscope attached to a Bio-Rad MRC 500 , s t '49 - confocal fluorescence imaging system. Conventional EM and immunoelectron microscopy were as described in ref. 23. Immunoadsorption and Gel Analysis. To isolate radiola- beled granule proteins, we used either rabbit anti-cytolysin or 4 rabbit anti-granzyme A (gifts of P. Henkart), followed by protein A-Sepharose as described (20). The latter antibody reacts preferentially with the larger of the two enzymes with Na-benzyloxycarbonyl-L-lysine thiobenzyl esterase activity found in NK granules (P. Henkart, personal communication) and can reprecipitate much of the 29-kDa polypeptide that is isolated with SBTI (data not shown). Therefore, this enzyme is most probably granzyme A (7). Some immunoprecipitates were with H digested overnight endoglycosidase (endo H; FIG. 1. Electron micrograph of RNK-16 cell showing the lytic endo-N-acetylglucosaminidase H, EC 3.2.1.%; Miles) at 30 granules. I, type I granule. Note that the dense core is surrounded by ng/ml, followed by a second digestion with endo H at 15 a thin membranous cortex (arrow). II, type II granule containing ng/ml for another 10 hr. Samples were then analyzed by multiple membrane profiles. Int, intermediate-type granule with a electrophoresis in 9-15% or 7-12% polyacrylamide gels with large type II cortex surrounding a type I core. M, mitochondrion; NaDodSO4, as described (20). PM, plasma membrane. (x24,000.) Downloaded by guest on September 25, 2021 7130 Immunology: Burkhardt et al. Proc. Natl. Acad. Sci. USA 86 (1989) pleiomorphic and contain numerous vesicular profiles. In- termediate forms are frequently seen. We used immunoelectron microscopy to determine whether the two most characteristic proteins ofNK granules in fact reside in the same organelle and, if so, in which one. Ultrathin frozen sections of RNK-16 cells were labeled with anti-cytolysin followed by protein A-colloidal gold. As shown in Fig. 2A, the label is primarily over the dense cores of type I granules. Type II granules and the cortical regions of type I granules are devoid of label. Tracks of endoplasmic reticulum (Fig. 2A, arrowheads) are labeled, but at a low density. Some labeling of the endoplasmic reticulum and Golgi is expected, since cytolysin passes through these its maturation below). FIG. 3. Double-label immunofluorescence ofRNK-16 cells using organelles during (see laser confocal microscopy. (A) Anti-cytolysin at 1:100 dilution fol- To localize serine proteases, we used biotinylated SBTI lowed by fluorescein isothiocyanate goat anti-rabbit immunoglob- and streptavidin-colloidal gold. As with the anti-cytolysin, ulin. (B) Biotinylated SBTI at 1:400 dilution followed by Texas the SBTI label is distributed throughout the dense core of red-streptavidin. The images shown are ofthe same optical section. type I granules and is absent from the cortical regions (Fig. Note that all the structures that are positive for cytolysin are also 2B). The specificity of the immunolabeling is demonstrated positive for serine proteases. by the lack of reactivity with nuclei, mitochondria, or the plasma membrane. It should be noted that although SBTI oligosaccharides. In addition to the cytolysin band, the detects all trypsin-like proteases in the cell, it labels only type tunicamycin immunoprecipitate consistently contains an- I granules. Since there are at least two distinct trypsin-like other polypeptide (80 kDa). This polypeptide may be the granzymes (7, 9), both must be in type I granules. As with binding protein (25) [or grp78 (26)], which is induced by cytolysin, type II granules are essentially devoid of label tunicamycin treatment and binds to various unglycosylated under all the conditions we used. This is not due to a technical proteins (26). problem such as inaccessibility to , since type II In pulse-chase experiments (Fig. 4C), the oligosaccha- granules can be labeled with a variety of other antibodies rides of newly synthesized cytolysin become resistant to (J.K.B. and S.H., unpublished data). Thus, we find that both endo H digestion with a til2 of 70-80 min. All the oligosac- cytolysin and the serine proteases reside in the same or- charides mature into the complex type, as indicated by the ganelle, the type I core. resistance of both mature bands (upper two arrowheads in The finding that serine proteases and cytolysin colocalize Fig. 4C) to endo H. This is evidence that cytolysin passes was confirmed by double-label immunofluorescence. In each through the trans-Golgi en route to type I granules. RNK-16 cell we could detect 10-20 granules by immunoflu- Granzyme A is also a glycoprotein. Newly synthesized orescence with anti-cytolysin antibody (Fig. 3A). Essentially granzyme A migrates as a 29-kDa band, which is reduced to every granule that was positive for cytolysin was also posi- 25 kDa after digestion with endo H (Fig. 4D). This is tive for serine proteases (Fig. 3B). consistent with the presence of a single asparagine-linked Asparagine-Linked Glycosylation of Cytolysin and Gran- oligosaccharide. One glycosylation site is indeed predicted zyme A. Metabolic labeling with [35S]methionine was used to by the gene sequences of both the murine and human follow the biosynthesis of cytolysin and the major serine granzyme A (8, 27), which are very homologous. Pulse- protease granzyme A. Mature cytolysin migrates as a doublet labeled granzyme A remains sensitive to endo H, even after of 67.5 and 70 kDa on NaDodSO4/PAGE (Fig. 4 A and B). 3 hr of chase (Fig. 4D), indicating that its glycan remains in When the addition of asparagine-linked glycans is inhibited the high-mannose form even when packaged into the lytic by treating RNK-16 cells with tunicamycin, cytolysin mi- granules. In this respect, the posttranslational processing of grates as a single band of64 kDa (Fig. 4B). This suggests that cytolysin and granzyme A differs. the two bands represent heterogeneity in asparagine-linked Phosphorylation of Mannose Residues on Granzyme A. glycans, probably the presence of either one or two complex Another difference between the posttranslational processing A

FIG. 2. Immunoelectron microscopic localization ofcytolysin and serine prote- ases. (A) Ultrathin frozen section of RNK-16 cell labeled with anti-cytolysin at a dilution of 1:30 and protein A-colloidal gold (9 nm). Gold particles are present over the core regions ofboth type I (1) and intermediate type (Int) granules but are absent from the cortical regions (arrow). Arrowheads delineate tracts of endoplas- mic reticulum, which are also lightly la- beled. M, mitochondrion. (B) A similar section labeled with biotinylated SBTI at a dilution of 1:140 and streptavidin- colloidal gold (10 nm). Label is present PM overthe type I core region (I) but is absent ..~ from the type II (11) granules and the type I cortex. The central density in the type II granule shown here is not a dense core but A 0 5pm O.5pn an artifact of the negative staining. PM, plasma membrane. Downloaded by guest on September 25, 2021 Immunology: Burkhardt et al. Proc. Natl. Acad. Sci. USA 86 (1989) 7131

ARNK B18 B Tun: - + the resulting monosaccharides by thin-layer chromatogra- A. phy. The radioactive label migrates as a single species, _ largely overlapping the mannose 6-phosphate standard, but 70 A 675 I 464 clearly distinct from glucose 6-phosphate, N-acetylglu- cosamine 1-phosphate, or Pi (Fig. 5C). We conclude that granzyme A bears a phosphomannose moiety, most likely i.:...... :.i- 0° 30 90' 180' mannose 6-phosphate. + + I_ + +r- Relative Hydrophobicity of Cytolysin and Granzyme A. ----:up: C Both cytolysin and granzyme A can be extracted without :- <70-w detergents (8, 21), but the fact that cytolysin can form pore 06 :; ~:.. -67.5 - _ complexes in target cell membranes suggests that cytolysin -64 (at least in its polymerized state) has hydrophobic properties. To assess the hydrophobicity of cytolysin and granzyme A, n Triton X-114 lysates of metabolically labeled cells were separated into aqueous and detergent phases as described _ ___* _*<~~~~~25 (20), and then cytolysin and granzyme A were isolated from each phase by immunoprecipitation. The lysis buffer and all subsequent solutions contained 2 mM EDTA to prevent oligomerization of cytolysin. Sixty-four percent of the radio- FIG. 4. Asparagine-linked glycosylation of cytolysin and gran- active cytolysin partitioned into the detergent phase, a hy- zyme A. (A) Anti-cytolysin specificity control showing immuno- drophobic character comparable to many membrane glyco- precipitation of metabolically labeled RNK-16 cells and an equiva- proteins. In contrast, 72% of granzyme A partitioned into the lent lysate ofthe hybridoma B18.8 (cytolysin negative). Arrowheads aqueous phase (n = 3; in a typical experiment, =20,000 cpm mark the mature cytolysin doublet. *, Contaminating actin. (B) were incorporated into each protein). Preliminary data indi- Cytolysin immunoprecipitated from RNK-16 cells labeled with me- cate that the reason for this differential hydrophobicity may thionine for4 hr in the presence or absence oftunicamycin (Tun). (C) be yet a third difference in the posttranslational processing of Maturation of cytolysin. Cells were pulse-labeled with [35S]methio- nine for 30 min and chased for the indicated times. After immuno- the two proteins, as cytolysin can be labeled with [3H]- precipitation with anti-cytolysin, half of each sample was digested palmitate (data not shown). The exact nature of this modifi- with endo H (lanes +). To better resolve the maturation intermedi- cation is not known. ates, a 20-cm gradient NaDodSO4/polyacrylamide gel was used. Arrowheads indicate the three glycosylation forms of cytolysin. (D) Maturation of granzyme A. Samples from the same experiment were DISCUSSION subsequently precipitated with anti-granzyme A and analyzed in the One major conclusion from the present study is that, in a rat same fashion. Arrowheads indicate the glycosylated and the degly- NK cell, cytolysin and granzyme A are packaged together in cosylated forms of granzyme A. The bands above and below are irrelevant nonglycosylated proteins. the dense matrix of type I granules. The dense core lytic granules also contain a chondroitin sulfate A proteoglycan, as ofgranzyme A and cytolysin is demonstrated by labeling with judged by subcellular fractionation and EM autoradiography phosphate. Granzyme A is phosphorylated, while cytolysin is (11, 28). It seems likely that, as is commonly observed in not. All the 32p; seems to be incorporated into oligosaccha- other secretory vesicles (3), the proteoglycan facilitates the rides, since it is removed quantitatively by digestion with condensation of type I granule proteins into the dense core. endo H (Fig. 5A). The phosphorylated oligosaccharides mi- The distribution of cytolysin described here for rat NK cells grate on a gel filtration column as a single peak the size ofthe resembles that of perforin in human CTLs (17), and the Man5GlcNAc intermediate (Fig. SB). The phosphorylated distribution of rat trypsin-like proteases resembles that of oligosaccharides all bind to the anion exchanger DEAE- granzymes D, E, and F in murine CTLs (18). All these Sephadex and are eluted as a single peak with 100 mM NaCl proteins are concentrated in the dense core parts of type I (data not shown). The behavior on both these columns granules and are not detectable in type II. suggests that the oligosaccharides bear a mannose 6- The physiological significance ofthis colocalization for the phosphate modification (24). To identify the phosphorylated natural killing process is still unclear. Although at least one moiety, we subjected the 32P-labeled oligosaccharides from serine protease, granzyme A, is secreted with cytolysin (6, the gel filtration column to mild acid hydrolysis and analyzed 10), its role in cytolysis has not been established. It has been

A B C endo H:- + 1 2 3

AL.

140 250 270 290 310 330 350 Fraction -Ori

FIG. 5. Modification of granzyme A with phosphomannose. (A) NaDodSO4/PAGE of granzyme A immunoprecipitated from 32PI-labeled RNK-16 cells before and after digestion with endo H. (B) Gel filtration of 32P-labeled oligosaccharides released from serine proteases by endo H digestion. The Bio-Gel P-4 column was calibrated with 125I-labeled bovine serum albumin (VO), Man7GlcNAc (M7), Man6GlcNAc (M6), Man5GlcNAc (M5), and nitroiodophenylacetyl (NIP). (C) Thin-layer chromatography of phosphorylated monosaccharides prepared from the peak shown in B. Fractions from the gel filtration column were pooled, concentrated, and subjected to mild acid hydrolysis. Equal portions of the labeled hydrolyzate were mixed with unlabeled standards and chromatographed on Silica G plates in solvent A. Lanes: 1, 32P, alone; 2-4, samples mixed with glucose 6-phosphate, mannose 6-phosphate, and N-acetylglucosamine 1-phosphate, respectively. The positions of the unlabeled standards are traced over the autoradiogram. Ori, line of sample application. Downloaded by guest on September 25, 2021 7132 Immunology: Burkhardt et al. Proc. Natl. Acad. Sci. USA 86 (1989) suggested that the granule serine proteases activate other will be very interesting to decipher the cellular control over lysis-specific proteins (7). If so, they are unlikely to activate the routing of each of these mannose 6-phosphate-bearing cytolysin before it is secreted, because the acidic pH of the proteins to the different parts of NK lytic granules. granules is unlikely to allow them to act (9). Indeed, we found no evidence for proteolytic processing of cytolysin during its We thank Dr. C. Reynolds for providing the RNK-16 cell line and biosynthesis. Dr. P. Henkart for his gifts of antibodies and for his continued The second major conclusion from our studies is that support. We are grateful to Dr. B. Kaufman (Department of Bio- although they reside in the same lytic granule, cytolysin and chemistry, Duke University) for help with carbohydrate analyses, granzyme A differ markedly in their biosynthetic processing. and to Dr. P. Matsudeira (Whitehead Institute, Massachusetts In- At stitute of Technology) for his hospitality in the use of the confocal least three posttranslational modifications distinguish the imaging system. This work was supported by Grants AI-08897 and two granule proteins. First, mature cytolysin bears complex CA-14236 from the National Institutes of Health. J.K.B. was sup- type endo H-resistant glycans, whereas the glycans of gran- ported by National Intitutes of Health Training Grant CA-09058. zyme A remain in the less mature high-mannose form. Second, Y.A. was a recipient of Grant JFRA-207 from the American Cancer granzyme A (but not cytolysin) bears phosphomannose resi- Society. dues. Third, cytolysin can be labeled with palmitic acid while granzyme A cannot. 1. Henkart, P. A. (1985) Annu. Rev. Immunol. 3, 31-58. The conversion of the oligosaccharides of cytolysin to the 2. Kupfer, A., Dennert, G. & Singer, S. J. (1983) Proc. Nati. complex form is good evidence that this protein is trans- Acad. Sci. USA 80, 7224-7228. ported through the trans-Golgi compartment. Since the ter- 3. Burgess, T. L. & Kelly, R. B. (1987) Annu. Rev. Cell Biol. 3, minal in 243-294. glycosylation enzymes reside the mid- and trans- 4. Millard, P. J., Henkart, M. P., Reynolds, C. W. & Henkart, Golgi (29), we interpret this modification to indicate that P. A. (1984) J. Immunol. 132, 3197-3204. cytolysin passes throughout the entire Golgi complex before 5. Podack, E. R., Young, J. D.-E. & Cohn, Z. A. (1985) Proc. reaching the type I granules. The persistence of endo H Nail. Acad. Sci. USA 82, 8629-8633. sensitivity of granzyme A is more difficult to interpret. It is 6. Pasternack, M. S. & Eisen, H. N. (1985) Nature (London) 314, unusual for a protein to traverse the Golgi complex without 743-745. any of its glycans being converted to the complex type. One 7. Masson, D. & Tschopp, J. (1987) Cell 49, 679-685. possibility is that its glycans are not substrates for terminal 8. Gershenfeld, H. K. & Weissman, I. L. (1986) Science 232, glycosylation enzymes, even when present in the trans-Golgi. 854-858. is that A 9. Henkart, P. A., Berrebi, G. A., Takayama, H., Munger, W. E. The more intriguing possibility, however, granzyme & Sitkovsky, M. V. (1987) J. Immunol. 139, 2398-2405. does not pass through the trans-Golgi on the way to type I 10. Takayama, H., Trenn, G., Humphrey, W., Jr., Bluestone, granules, but rather is diverted to the granules from an earlier J. A., Henkart, P. A. & Sitkovsky, M. V. (1987) J. Immunol. compartment. 138, 566-569. Granzyme A must be transported to at least the cis-Golgi, 11. MacDermott, R. P., Schmidt, R. E., Caulfield, J. P., Hein, A., since the enzymes responsible for the mannose 6-phosphate Bartley, G. T., Ritz, J., Schlossman, S. F., Austen, K. F. & modification apparently reside there (30). Moreover, the Stephens, R. L. (1985) J. Exp. Med. 162, 1771-1787. phosphorylated oligosaccharides have been trimmed to the 12. Frey, T. H., Petty, R. & McConnell, H. M. (1982) Proc. Nail. Man5GlcNAc2 intermediate, showing that they have encoun- Acad. Sci. USA 79, 5317-5321. 13. Orye, E., Plum, J. & DeSmedt, M. (1984) Histochemistry 81, tered the cis-Golgi mannosidases (30). Thus, granzyme A and 287-290. cytolysin could be sorted from one another as early as the cis- 14. Zucker-Franklin, D., Grusky, G. & Yang, J.-S. (1983) Proc. or mid-Golgi. Although most proteins are sorted in the Nati. Acad. Sci. USA 80, 6977-6981. trans-most elements of the Golgi complex (31), there is also 15. Zaguri, D. (1982) Adv. Exp. Med. Biol. 146, 149-163. evidence for sorting of lysosomal enzymes at the cis-Golgi 16. Neighbour, P. A., Huberman, H. S. & Kress, Y. (1982) Eur. J. (32). It has not yet been determined whether granzyme A is Immunol. 12, 588-595. transported through the trans-Golgi en route to the granules. 17. Groscurth, P., Qiao, B.-Y., Podack, E. R. & Hengartner, H. The finding of a mannose 6-phosphate modification on (1987) J. Immunol. 138, 2749-2752. granule serine proteases is very intriguing. Additional studies 18. Jenne, D., Rey, C., Haefliger, J.-A., Qiao, B.-Y., Groscurth, P. & Tschopp, J. (1988) Proc. Natl. Acad. Sci. USA 85, 4814- are necessary to determine whether this modification is 4818. actually used as a granule targeting signal in the manner that 19. Reynolds, C. W., Bere, E. E., Jr., & Ward, J. M. (1984) J. it serves as a lysosomal targeting signal for other proteins. If Immunol. 132, 534-540. mannose 6-phosphate is indeed a targeting signal for gran- 20. Argon, Y. & Milstein, C. (1984) J. Immunol. 133, 1627-1633. zyme A, it will have interesting implications. First, the use of 21. Reynolds, C. W., Reichardt, D., Henkart, M. & Millard, P. mannose 6-phosphate-mediated targeting will add a func- (1987) J. Leukocyte Biol. 42, 642-652. tional similarity between lytic granules and lysosomes. This 22. Takatsu, K., Sano, Y., Hashimoto, N., Tomita, S. & Hamaoka, will be in addition to the other known common properties, of T. (1982) J. Immunol. 128, 2575-2580. acidic pH and the presence of "classic" lysosomal enzymes. 23. Burkhardt, J. K. & Argon, Y. (1989) J. Cell Sci. 92, 643-654. will that there are at least two 24. Distler, J., Hieber, V., Sahagian, G., Schmickel, R. &Jourdian, Second, it indicate biosynthetic G. W. (1979) Proc. Nail. Acad. Sci. USA 76, 4235-4239. routes for dense core granule proteins, since some of them, 25. Bole, D. G., Hendershot, L. M. & Kearney, J. F. (1986) J. Cell like cytolysin, do not use the mannose 6-phosphate signal. Biol. 102, 1558-1566. Another reason to suspect more than one granule targeting 26. Kozutsumi, Y., Segal, M., Normington, K., Gething, M.-J. & mechanism is the heterogeneity of glycosylation observed Sambrook, J. (1988) Nature (London) 332, 462-464. among the six murine granzymes (7). One ofthese, granzyme 27. Lobe, C. G., Finlay, B. B., Paranchych, W., Paetkau, V. H. & C, is unglycosylated (7, 18) and would therefore not be Bleackley, R. C. (1986) Science 232, 858-861. expected to use mannose 6-phosphate mediated targeting. 28. Dvorak, A. M., Galli, S. J., Markum, J. A., Nabel, G., Der Our data point out yet another level of complexity in Simonian, H., Goldin, J., Monahan, R. A., Pyne, K., Cantor, H., Rosenberg, R. D. & Dvorak, H. F. (1983) J. Exp. Med. 157, sorting granule proteins. Previous work had established the 843-861. subcellular distribution of a number of lysosomal hydrolases 29. Roth, J. & Berger, E. G. (1982) J. Cell Biol. 92, 223-229. in NK cells (12-14). Although all these enzymes presumably 30. Kornfeld, R. & Kornfeld, . (1985) Annu. Rev. Biochem. 54, bear phosphomannose, they are present in type II granules 631-664. and the cortical regions of type I granules, a distribution 31. Griffiths, G. & Simons, K. (1986) Science 234, 438-443. opposite that of the similarly modified serine proteases. It 32. Brown, W. J. & Farquhar, M. G. (1984) Cell 36, 295-307. Downloaded by guest on September 25, 2021